Modelo de learning analytics basado en métricas emocionales para aplicaciones móviles de aprendizaje en segunda lengua

Abstract

This document presents the results obtained from the implementation and application of a learning analytics model based on emotional metrics, which exposes two main components. The first component is focused on the application of visualization analytics techniques and the second is oriented on the application of emotional clustering techniques of variable length time series. This model consumes a data set called Emotional Data L2 Interaction (EDaLI), which was also built in this study. This dataset features emotional time series captured during the interaction of a group of 19 people with four Portuguese lessons from the Babbel second language learning app. Additionally, this data set is annotated, specifying the activities carried out by the participants, second by second during their interaction. The results obtained present two perspectives, the first oriented to visualization analytics, which allows us to observe and understand the emotional behavior and the results obtained by the participants at the group and individual level and in this way generate initial recommendations. The other perspective is focused on emotional clustering, where models are obtained that, based on metrics such as the Silhouette, Davies-Bouldin and Calinski-Harabasz index, present a medium-high level of cohesion, compactness and separation in the groups generated, from the emotional data analyzed